Local file and format I/O

traveler.io is the place to find tools for reading data from a local filesystem into memory for use by Traveler. It is a fundamental part of the normal model workflow and groups tabular readers and format-specific skim loaders without mixing file operations into Traveler’s in-memory table and store abstractions.

Tabular data

Use read_table() to load CSV or Parquet data into a Polars DataFrame. The reader infers the file format from its suffix and can select, rename, derive, sample, and schema-cast columns while the data remains in Polars. Convert the prepared result explicitly with traveler.tables.JaxTable.from_polars() when it is ready for model calculations.

Skim formats

OmxSkims loads selected Open Matrix (OMX) matrices and groups time-period matrices for conversion to a JAX dataset. The lower-level OMX interface reads and writes the OMX container itself. ZarrSkims provides the corresponding adapter for Zarr-backed skim data.

These loaders are format adapters. Skim lookup during model execution remains in traveler.skims.

Fetching remote example and test data is intentionally separate. See Remote asset utilities for the optional traveler.assets utilities.